Genetic Algorithm for Decentralized PI Controller Tuning of a Multi-Span Web Transport System Based on Overlapping Decomposition

摘要

The use of multi-span web transport systems often requires dedicating a particular effort for defining a control system able to protect the integrity of the web. In web handling processes there is a clear need for controlling web velocity and tension. As for any control problem, the best results are achieved when there is a clear understanding of the controlled process. Starting from a new mathematical model called Transfer Matrix Model for a more accurate description of a web transport system, an optimization study for identifying parameters of the PI controllers for a decentralized multiinput and multi-output system (MIMO) has been proposed. The possibility of using PI controllers for each section is attractive considering that an overlapping system decomposition may permit to take into account the mutual interaction of the neighbor sections reaching specified performance objectives. Finally, by using a nonlinear interpolation of the trend of a preliminary database of values obtained through genetic algorithm, a self-tuning strategy is proposed to estimate optimal PI parameters under certain conditions, avoiding the long identification process and making the system flexible and adaptive. Simulations and experimental results validate and illustrate the effectiveness and the simplicity of the proposed method by considering several different set-points.